September sessions added new worn recordings to the acquisition record. On 24 September, two 60-second baselines completed with no reported missing EEG packets, timestamp anomalies, or notification queue drops. A separate 30-second fresh-readiness recording also completed. These observations concern transport and recording continuity; they do not establish clean EEG.
The guided protocol required all nine candidate signal windows to pass its existing readiness checks. The two completed assessments on 22 September passed two and zero windows; the 24 September assessment passed four. Each session halted before labeled task trials. These counts are readiness outcomes, not decoder accuracy, and the higher count does not establish improved control.
Two further stationary baselines were completed and operator-accepted on 25 September. Their saved quality reports showed substantially more 50 Hz interference than the 24 September baselines, particularly at the right temple and right mastoid. No fresh-readiness assessment or task trial followed in that session. The physical cause remains unverified; accepting a baseline records its condition rather than certifying clean neural signal.
Offline review now compares the current readiness rules with a clearly labeled hypothetical change on saved data. It cannot authorize a participant trial or alter a live profile. Denoising comparisons likewise preserve the original recordings: a quieter trace can also remove neural information, so lower amplitude alone is not evidence of better decoding.
The next evidence step is to characterize the interference, establish usable signal on fresh recordings, and then collect separately labeled development and held-out trials. Reliable command accuracy, idle errors, decision latency, and repeatability remain unvalidated.